Iteration 82 — H-044 Congruence EXCLUDED KILL #24 · THE ORDERING LEG'S FIRST KILL · EXAM KILL

In plain words: this candidate reads whether the FACTOR STRUCTURE of the market — which hidden themes drive which features — rearranges between regimes. It became the first candidate since the distance cousins to clear the door law: its regime-to-regime drift reading genuinely exceeded what re-shuffled blocks produce. Then a subtler control exposed what that clearance was made of. The exam also asked: does the factor structure measured on two halves of the SAME regime agree better than structures from two DIFFERENT regimes? It must — otherwise the "drift" is just measurement wobble. It did not: same-regime halves disagreed MORE (agreement 0.55–0.59) than different regimes did (0.63). With 125 features, ~1500 rows, and 28 factors, the loading estimates are so noisy that estimation variance — not regime structure — dominates the reading. The honest nuance is on the record (halves have half the data, by design — but several campaign windows are smaller than these halves, so a reading that can't survive half-window noise can't serve the boundary). The boneyard prescribes the fix for any successor: a variance-controlled loading estimator, and the ordering leg first.